What do folks think about the need for an additional task framework (e.g. Invoke, Rake, Make, etc.) in repositories managed via Pants? Do you use a task framework for custom tasks, or do you bake them into Pants via plugins?

If possible it's better to write plugins, because then you get the right invalidation, caching and concurrency semantics. If you shell out to something like Make, you have to know how to capture and encapsulate the state differences it causes, and that can get really tricky if trying to do it in a general-purpose way.

Roger that. My current use case is for executing Copier (same idea as cookiecutter) to generate a directory structure from a template and wrap it with some default parameters.